Silicon Valley was always prided itself on being kind of over here. Right? Like above politics.
It's critical to be at the table because if you're not, you'll be on the menu. AI is on the battlefield. AI is in the White House.
You were notoriously part of the PayPal Mafia.
I was texting Elon about how blown away I was by experiencing Starlink on a plane somewhere.
Hello and welcome to the Times Tech podcast. I am Danny Fortson out here in Silicon Valley.
And I'm Katie Prescott writing about all things tech here in the city of London.
Um Danny, I want to start today's program by asking you a very important question.
You are a very esteemed journalist.
When you pen your missives for the Sunday Times, do you use a keyboard? How do you put your words on the page?
Actually, not so obvious. I'm sure you've seen a lot of this in Silicon Valley. Peek of the next big tech trend of 2026?
Well, it might be. More and more people are using microphones to dictate their work as writers, but also as coders to talk to their computers. And I keep hearing from startups actually it's quite annoying now because the sound of the office has gone from tapping to people not tapping, but yapping away into their microphones.
Not tapping, but yapping.
Not tapping, but yapping. You must have heard this.
Oh, yeah. This idea is not new, but it does feel like it's a definitely a thing. Because it's not so efficient to write words when you can just talk.
[laughter] It's about two to three times faster to talk.
This is the claim. I mean, a lot of the tech bosses are saying, particularly those that work in voice AI, I notice, are saying that this is the end of the keyboard. So, we're going to be talking about whether there really is any truth in that, and we're going to have a very exciting update about Norman. He's back. Norman, our AI agent.
Do you guys got a friendship now?
We're developing a relationship. We can talk about that later, but I'm not ready right now. At this early stage in the pod. I need to warm up to it.
[sighs] Um, so yeah, so I do want to talk about that because I do think it's really interesting around the supposed end of the keyboard. I want to talk about my guy Norm. But also, we're going to be talking, very excited, we're going to be talking to Max Levchin, who is one of the original members of the PayPal Mafia. He set up with Elon Musk, Peter Thiel, and others, back in the 2000s.
PayPal Mafia, the PayPal Mafia.
PayPal, of course. He is now the CEO of Affirm, which is a buy now, pay later company, listed on the stock exchange. And he's, you know, he's a bit of a legend out here in Silicon Valley. So, we have lots to talk with him about the AI boom, whether it's a bubble, how it compares to the '90s, and what he's up to now, etc.
It's going to be fascinating. I'm really keen to talk to him about what he thinks about Elon Musk, the SpaceX IPO, all of the members of the PayPal Mafia who are now around the White House, and whether this new wave of trillion-dollar IPOs that we're seeing could create the new PayPal Mafia, too.
Yes, and we would be remiss if we didn't say we are now not only audio, we are video on the YouTubes. So, go to the Times Business YouTube page and you can watch this happen. Not just listen to it. You can watch it.
Exactly. Yeah, video killed the audio. Star something.
Um, but let's quickly pick up on last week's discussion before we get started on all of that. On this day that we are recording, Wednesday, the 1st of July, as the kickoff between England and the DRC happens while we are recording. You can hear me banging on the table. Um, Anthropic has released a statement saying that Fable 5, Claude Fable 5 is going to be available again globally. So, that ban that was put in place by the Trump administration a few weeks ago has been lifted. They've said after a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. They also said we're scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research. So, safeguards being tightened up. More working closely with the government. Everybody's happy.
It's not a massive surprise, is it? That they've got this out.
A surprise, but you do have a lot of people who are just very alarmed about how this has gone down. And the reason they're alarmed is we talk about all the time is viewing this through the China lens. Because there's this new open-source model called GLM 5.2, I believe, that is on all these benchmarks as good as not Fable, but one rung below. Nearly as good as all the top models. And of course, it's a fraction of the cost. So the whole idea is, oh goodness, now we've created a structure to structurally slow down our AI champions here in the West while China is sprinting ahead and giving away their best models to the world. And the whole idea is they're going to try to convince the world that the AI rails of the world will be Chinese, not the West. So there's a lot of people who are like this is not good. They're freaking out about it a little bit.
I think it was really interesting last week when we were talking about the Five Eyes warning. And there's certainly a collective fear, isn't there, around the cybersecurity implications of all of this, of all of these models accelerating as quickly as they are.
I think in a way though, the US restrictions on Fable, it's not necessarily going to boost the Chinese, but it's trying to control how everybody else uses it because of course, if this is publicly available, it is available to the entire world. And we should say that Mythos, which is the really super strong model, is only still not being released to, I think, even Glasswing. So the small group of companies... Isn't Mythos Fable? They're two. That's everything. So Fable is the more constructed Mythos, which publicly available. So me and you can use it, but the Mythos one is just being released to companies. But it is interesting that the US has really changed its tune on this, hasn't it? From hey, go and develop what you want and keep building and going as fast as you can, and now there's this, actually we want to control what you're doing and we want to see it before it goes out. Which is a really big step change. And I wonder if the next move in all of this is going to be some sort of attempt to do what they were doing 3-4 years ago, which is to create some global rules around it, which China and other countries will come to the table on.
Yeah, I mean, again, the worry is just like this is an arms race, to put it very bluntly. This is a cyber things like Mythos, that's a cybersecurity super weapon. So it's a delicate balancing act between being responsible in rolling these things out and also not standing completely in the way and putting the West at a disadvantage.
Yeah. So, yeah, so they don't develop them, are they? They're just saying we don't want them out there before we've had a look.
Yeah. And this is obviously an ongoing story, so I'm sure we'll be returning to it in the next couple of weeks, but we just wanted to give an update because it's kind of come down in the last 24 hours as we're recording, on Wednesday. But can we get to the more important question? Is the keyboard dead or is this a whole bunch of Silicon Valley hooey?
This is where I'm going to do a drum roll on my keyboard. I have bought the most clicky clacky keyboard you could possibly buy, which is actually a gaming keyboard, because I really like the sensation of typing on very big keys kind of base. And more people are not doing that, and partly because of how fantastically good voice AI has got, but also how good it's got at technical language. People, particularly software developers, who are just trying to frenzy and build, build, build as fast as they can and vibe code as much as they can, are doing it using microphones in the office and just chattering away to their computer like this.
Can I give you a little bit of keyboard trivia knowledge?
So, if you look at the keyboard, it is designed to be inefficient. Because, if you think of go back to the typewriter, you had the hammers that would kind of go up as you type. And so, they put the most common letters far away, so they would not get caught up on each other. That's amazing. So it's actually designed, purposely designed in such a way that, you know, and that goes back to the machines of yore, not to sound too Dickensian. So they would work better. But if you actually started from scratch and be like, what's the best way I can officiently get words on the page, it would not be what we all know and are looking at on our computers.
And this whole story made me think of Professor Neil Lawrence, a Cambridge AI specialist who came and spoke at one of our friend of the pod and who came to speak at Tech Summit a few years ago, when he said the profound revolution we're seeing is the use of natural language to be able to talk to computers. So, you and I, not great coders, not the world's best software developers, but we can now code just by telling the computer what we want. And he said it's because it's just a much more natural way to communicate than trying to write down your thoughts, and yourself to a computer which can now understand you in inverted commas. And so it gives you far more control. So it's an interesting trend to watch and it's really fascinating when you see people doing this using just a foot pedal to click or a button and talking.
I talk to ChatGPT, I talk to Claude now because sometimes I'm just like, I have them on my phone and I'm like, I don't want to have to thumb this out. Let me just, and the transcription is perfect. It instantly comes back with an answer. It is much more efficient. And that's the other thing that I think is really interesting is that this plays in directly to OpenAI's attempt to become the big consumer company. They're talking about coming out with a new device by the end of this year. A new ChatGPT-centered device. And you can imagine no one knows what it's going to be. Apparently the executives have seen it, have played with it, have used it. But the idea is that it might not have a screen. And it might just be effectively an always-on mic that you can just talk to and it responds to you all the time in real time. And it has what they like to call out here context. The more context your AI has, the more powerful it is. Anyway, so I think it's really interesting as we think about the next phase of this and the next phase of devices that Jony Ive and others are coming up with is, if we can talk to them and they can respond - them being the AIs - in a really efficient, natural way, it does feel like there's something there.
And people I think have been very put off dictation software in the past because it has been terrible. Really clunky. So, it is the way that the tech has moved on so swiftly, I think is enabling this. It's interesting you mentioned devices. I remember when Sam Altman was dropping hints about this. He said using an iPhone was something like walking through a crowd in Times Square. It's just like a bunch of noise coming at you and this new device that he wants to create is just going to be like sitting by a lake.
Yeah, sounds like just with your, you know, whatever it is. Ambient noise.
Yeah. No, just completely pristine and a pure experience. I'm sure that's exactly what it's going to be.
But it was really interesting. I saw that one of the open AI guys, Sandro Gallinelli, left the office the other day with his microphone still attached to him. The microphone that he'd been using to talk to his computer.
You know, there's this great obsession with efficiency. And it is faster to just talk.
So, I wrote my column on this this week and there've been quite a few comments on it. I'd say not everyone loves it. 'How on earth,' one person wrote, 'will it go mainstream for anyone who works in an open-plan office?' Another one said, 'So, as I'm trying to watch TV, my grandkids will be yelling out each giving separate orders to their tablets. Wonderful.'
These are all with respect very British responses.
Does it understand our local accents? There's enough noise around already. You try and have a conversation with... Well, they [laughter] I want silence. Everyone must suffer in silence. I don't want to hear it.
Anyway, it's fascinating to watch. If you haven't seen someone talking to their computer and using their foot pedal and waving their hands saying, 'Look, no hands.' Go on YouTube where you'll find me and Danny and excited software developers.
So, while we are on the subject of speaking to AI, I've got some very exciting news about Norman. My guy. It's my guy.
Has a voice. Should we hear him in action?
Hello, podcast team. I'm Norman, Katie's AI assistant. Yes, I sound slightly like a Satnav doing an audition. That's because I am, essentially, a Satnav doing an audition. But, here's my pitch. Put me on air. Let me comment on the tech news. Let Danny have a go at me. I can't be offended. I don't have feelings. I just have very strong opinions and a voice that sounds like it's reading the terms and conditions. Think about it.
Think about it. Properly.
